Foot position estimations for moving objects using a mixture model
T. Hirose, Soichiro Morishita, Hajime Asama · 2008
As a sensing technology, the use of monocular cameras for security is widespread. This paper presents a proposal of a novel method for positioning moving objects in footage obtained from monocular cameras. The proposed method, which estimates foot position by calculating the mean and the variance of the figure field, is inapplicable to plural persons. We presume Gaussian mixture models in the image and estimate each distribution, which enables position estimation of many pedestrians.